Sora 2 Multi-Character Interaction Failure — Refund Guide
Technical Classification
Multi-Agent Topology Collapse
Multi-Agent Topology Collapse occurs when Sora 2's character-conditioning model fails to maintain identity and body separation across multiple subjects in close-range interaction. Bodies fuse where they should overlap, faces swap features between subjects, or one character's clothing migrates onto the other. The failure is most pronounced in hugs, handshakes, dance partners, fight choreography, and any scene where two subjects share a contact region.
How to identify this failure
- ✕Two characters' bodies fusing where they should overlap
- ✕Facial features swapping between subjects
- ✕Clothing migrating from one character to another
- ✕Limbs from different characters interpenetrating
- ✕One character's identity collapsing into the other
Real generation examples
Prompt used
"Two old friends hugging at an airport arrivals gate"
Failure observed @ 0:02
Shoulders fused at contact, one face features migrated to the other by 0:02
Prompt used
"Couple slow-dancing in a candlelit room"
Failure observed @ 0:01 → 0:04
Hand-holding region collapsed into single fused hand at 0:01; identity drift at 0:04
Documentation strength
If you need to escalate
VERY HIGH — Sora 2 markets identity preservation across subjects; multi-agent collapse is a recognised flagship-feature defect.
AVA is a pre-purchase prevention tool, not a post-purchase recovery tool. Platforms generally do not guarantee credit refunds for output-quality failures; goodwill credits are at each platform's discretion. The strength rating reflects how well-formed your support ticket can be, not a promised outcome.
Prevention + documentation steps
- 01
Score your prompt before you generate
Run your prompt through AVA's pre-flight scoring against the Multi-Agent Topology Collapse pattern. Green light = generate. Yellow/red = rewrite using the suggested fix before you commit credits.
- 02
Capture Generation ID + timestamp if it failed anyway
Find the Generation ID in the URL or share link. Note the exact time when the Multi-Agent Topology Collapse first appears (e.g. "failure first visible at 1.2s"). Timestamped evidence is significantly stronger than a general complaint.
- 03
Use the correct technical term in your support ticket
Describe this failure as "Multi-Agent Topology Collapse". This term maps to a recognised internal workflow in the support system and routes the ticket to the right team.
- 04
Submit via the correct support channel
Runway has no direct email intake. Pro+ plan: open the in-app AI Assistant (help widget bottom-right of app.runwayml.com), describe the failure with the technical term, attach evidence. Free/Standard plan: human support isn't available — your channel is Discord #community-help with @On Call - Moderators.
Frequently asked questions
Does OpenAI refund Sora 2 multi-character failures?
Yes. OpenAI support treats multi-agent topology failures seriously because identity preservation is a Sora 2 marketed capability. Cite the technical term, document with frame stills, and submit through the Sora support portal.
Why does Sora 2 merge characters?
The character-conditioning model shares latent capacity across all subjects in a scene. When contact regions enter the same patch tokens, identity features bleed between subjects.
Which Sora 2 prompts are highest risk?
Hugs, handshakes, fight choreography, dance partners, and any close-contact multi-subject scene. AVA flags multi-character interaction prompts for elevated risk scoring.
Score your prompt
Score your prompt against this failure mode in 30 seconds
Paste your prompt and the platform you intend to use. AVA returns a red/yellow/green score against this specific failure mode plus a concrete rewrite if the risk is high.
AVA Pro · founders' round
$50 for 6 months of unlimited scoring across all failure modes + personal failure-history dashboard. Locks in $13/mo grandfathered after.
Related failures across models
If you’re seeing this failure, you may also encounter these on other models:
Physics
Fluid inversion, gravity violations, object interpenetration, and impo…
Semantic
Sora output ignores or contradicts explicit prompt instructions — wron…
Provenance
C2PA provenance watermark or Sora logo persists in the visible frame o…
Facial
Sora output shows face morphing, identity inconsistency, or feature di…
OCR-Coupled
Sora output contains visible text that is misspelled, illegible, or re…
Skeletal
Sora output contains anatomically impossible human or animal bodies — …
Pick a different tool for Sora failures
Some prompt shapes will keep failing on Sora. Routing those shots to a different vendor is the cheapest fix.